Mobile to Models and Leading Design with Craft

Most design leaders right now are treating AI the way they treated mobile in 2008: delegating it, outsourcing judgment, waiting for patterns to settle. That worked then because the breakthroughs were spatial. It will not work now because the breakthroughs are behavioral. And behavioral systems require a different kind of leadership entirely.

We are in a shift. And most leaders have not moved yet.

Stable-era leadership that was

For much of the past decade, leadership was about taking mature materials and making them work at scale. Mobile had moved from experimentation into conventions. The web had hardened into design systems. The work was no longer about invention. It was about embedding patterns, coaching teams, and scaling reliably.

In that era, presence mattered. Empathy mattered. Process mattered. These qualities anchored teams while the material matured. They remain important today, but they are no longer sufficient.

When the material evolves, leadership is called to action again.

The new material

Software used to be deterministic. You designed a flow, shipped it, and it behaved the same way every time. The design surface was interface, structure, and usability.

Models are different. They are adaptive and non-deterministic. The same input can produce different outputs depending on phrasing, context, or temperature. They flex. That flexibility is the point, but it introduces complexity. Cost is now part of the design surface. Context windows expand and contract. Outputs drift in tone and style based on invisible signals. Personalization happens in prompts that rewrite themselves and systems that adapt behind the scenes.

This is not the vocabulary of tabs and flows. Unless leaders become fluent in it, they cannot guide what these systems can or cannot do.

Behavior is a harder dimension than space

When mobile first arrived, the breakthroughs were about where things lived. Where navigation sat. How surfaces reacted to touch. We debated top versus bottom navigation. We turned cameras into scanners. These were spatial problems. You could see them, test them, iterate directly.

The breakthroughs with models are about how systems behave over time. The system changes tone. It shifts personality. It adapts as people interact with it. You cannot see the full arc of that behavior in a single screen or flow.

That invisibility is the challenge. Leaders are no longer choreographing surfaces. They are shaping behavior without being able to fully control it. This requires a different posture. Not the steady hand of a systems-builder. The experimental rigor of a craft-builder.

What failure looks like

When design leaders do not understand the new primitives, the organization cannot define what good looks like.

Consider a customer support experience powered by AI. It answers questions correctly. Resolution rates are strong. Speed is good. But users describe it as “robotic” or “frustrating to talk to” even when their problem gets solved. The issue is not accuracy. It is tone, pacing, and empathy: qualities that were never part of the evaluation rubric.

Because leadership did not recognize these as critical dimensions, the organization never built systems to measure or improve them. Product reviews focus on resolution rates. Design reviews focus on layout. QA tests correctness, not whether responses feel human. No one asks: Does this sound empathetic or transactional? Does it acknowledge frustration or just provide information?

These questions are not part of the process because leadership has not identified them as part of the new standard. Without that standard, teams optimize for what gets measured (speed, accuracy, completion) and the dimensions that shape user trust go unnoticed and unfixed.

AI fundamentally changes what “good” means. Resolution speed still matters. But now emotional resonance, behavioral consistency, and contextual awareness matter just as much. If leadership does not establish the new standard, the organization will keep optimizing for the old one. Work will ship. Metrics will look fine. But the experiences will feel hollow.

The machinery resets

The machinery is everything that defines and measures quality: design reviews, evaluation criteria, the principles that guide what gets shipped. These systems were built for deterministic software. They ask: Does this make sense? Is it easy to use? Does it align with our visual language?

Those questions still matter. But they are no longer sufficient. Evaluation cannot be based on fixed flows when flows change with each pass. Reviews cannot focus only on layout when tone and behavioral consistency now shape trust. The principles that guided deterministic software (predict the outcome, control the output, ensure repeatability) do not apply when the system adapts and behaves differently based on invisible signals.

The bar for quality has moved. If the systems that enforce quality have not moved with it, the organization will keep optimizing for the wrong things.

The responsibility sits with leadership. Not to dictate every decision, but to recognize the new dimensions of quality and establish the systems that guide teams toward them. If leaders are not close enough to the material to see what matters, they cannot guide the organization. And if they cannot guide it, no one else will.

Lead with craft

It does not mean hovering over work or tightening process. It means immersing yourself in the material until you develop the judgment your teams will rely on.

Get hands-on. Use the tools yourself. Build something, even if small. Play with prompts. Push boundaries. See where they break. Judgment is built by feeling the material resist, surprise, or collapse under pressure. Prototypes sharpen intuition faster than theory.

Understand evaluation. Learn what rigor looks like in this material. How do you know if tone is consistent? How do you test whether the experience holds up under varied phrasing or at the edges? What does graceful degradation look like? If you cannot evaluate the work, you cannot establish standards.

Know what models are good at. Spend time understanding what different models do well. Some handle empathy better. Some are better at structure. You do not need to become an expert. You need enough fluency to have an opinion when your team says “we are using this model for this task” and ask whether it is the right fit.

Codify what you learn. Translate discoveries into shared patterns and principles. Turn them into playbooks your team can rely on. Make the invisible visible.

Spend more time than most. Depth is built by living closer to the material than others. That depth is what lets you coach from fluency, set sharper expectations, and raise the standard for the craft.

This commitment doesn’t look optional. We’re firmly in the AI era and the work is only going to trend up from here.

In stable times, leaders can set systems and step back. When the material shifts, leaders have to step in. Not to micromanage. To set the standard.

Leadership now is about shaping the new craft through proximity, through judgment, through translation back into principles your teams can carry forward. It is about being the person who knows what good looks like because you have done the work to find out.

If you are not close to the material, you can coach on process. You can advocate for teams. But you cannot lead on craft. And craft is what this moment requires.

This is the responsibility. Not someday. Now.

With that, back to work.